It is true that in most of the arXiv the author has to supply the journal of publication, and that most authors don't do this. However the SLAC SPIRES service at Stanford supplies this data to the arXiv in high-energy physics. For example, if you look at the first 100 articles in the hep-th archive in December 1998,
http://arxiv.org/list/hep-th/9812 you will see that 77 of the 100 have journal-ref fields. I just did a cursory review of the other 23. At least 4 of these are published in journals but were missed by SPIRES. Of the other 19, 11 are labelled by the authors as conference proceedings or invited lectures, and 2 are Ph.D. theses. Thus at least 94 of the 100 have been blessed by some form of peer review. Now I know that there is a damned-if-you-do, damned-if-you-don't argument against the arXiv: if an arXiv article is published, it's theft, and if isn't published, it's crap. However, in high-energy theory the publishers know full well that their articles are in the arXiv, and that they are going to stay in the arXiv. They are still willing to publish these articles, presumably because otherwise they wouldn't have any high-energy theory articles to publish. Besides, just because an article isn't published, that doesn't prove that it's crap. It could be a confident work by a superstar who is no longer on the publish-or-perish treadmill. (math.GT/9712268 is one example.) According to the recent New York Times article on the arXiv, http://www.nytimes.com/2001/05/01/science/01ARCH.html most high-energy theorists are not only awed by the average quality of hep-th, but also don't mind that it accepts a few never-to-be-published as well as obvious oddball submissions. You can view it as egalitarianism. I can also address George Lundberg's question about how readers are to be warned about arXiv articles that aren't and shouldn't be published. Or, more often, arXiv articles that shouldn't be published but are anyway. It may be different in medical research, but in math and physics most mediocre papers are sufficiently boring that almost no one reads them or cites them. (In math this happens to very good papers too, unfortunately.) After a month or so even green readers generally don't find these papers even if they are in the arXiv. A few papers are both provocative and mediocre, and some such papers are criticized in other arXiv papers. E.g. if I understand things correctly quant-ph/0003036 discredits quant-ph/9806088, even though the latter was published in Physical Review Letters. In mathematics it is less common for people to criticize other people's papers but more common for people to criticize their own. Most of us live by a code of certitude that would be impossibly strict in many disciplines. This is not necessarily because we're more honorable people; rather the subject matter is relatively precise. It is consequently devastating to your reputation if you refuse to admit error. For instance a recently withdrawn math arXiv article, math.NT/0106267, is perfectly correct and interesting; the authors retracted it solely because the result was previously known. -- /\ Greg Kuperberg (UC Davis) / \ \ / Visit the Math ArXiv Front at http://front.math.ucdavis.edu/ \/ * All the math that's fit to e-print *